Predicting household air conditioners’ on/off state considering occupants’ preference diversity: A study in Chongqing, China

Abstract

The input setting of an air conditioner (AC) usage schedule has a significant influence on the results of the bottom-up simulation model of household AC energy use. To set more accurate inputs when estimating the AC energy consumption based on the bottom-up simulation model, it is essential to establish a prediction model for the AC on/off state. However, due to the limitation of the data collection of household AC usage, the occupant diversity of household AC usage behavior could not be considered fully in existing research. This study aims to establish a prediction model of the AC on/off state combined with occupant behavior diversity on AC usage preference, based on the monitoring dataset of 1,274 ACs in Chongqing, China. First, clustering analysis was employed to identify the occupant thermal and schedule preference diversity on AC usage. Three typical thermal preference patterns were identified for wall-mounted AC and four for floor-standing AC. Three typical schedule preference patterns were discovered for both the wall-mounted AC and floor-standing AC. The results of preference diversity, as well as other environmental and time parameters, were used as inputs for the AC on/off state modeling. In this process, XGBoost is used to establish the prediction model. The proposed model shows good performance for the majority of AC samples in the testing dataset. For wall-mounted AC, in the testing dataset, the F1-score is above 0.7 for approximately 70% of the total AC samples; for floor-standing AC, the F1-score of approximately 50% of the total AC samples is above 0.67. The results and methods in our study could be helpful for improving the input setting of AC schedules in a bottom-up simulation model of household AC energy use.